Labelled Subgraph Query Benchmark – A lightweight benchmark suite focusing on subgraph matching queries. Note: This is a microbenchmark for system developers and not an official LDBC benchmark.
Currently, if any query crashes, the measurements stop for the given SF.
To work around this, every query should run on a freshly started DB. This considerably slows down the benchmark for small SFs but for large SFs, the effect is more limited (maybe adding 9x0.5 min ~ 5 minutes in total).
Currently, if any query crashes, the measurements stop for the given SF. To work around this, every query should run on a freshly started DB. This considerably slows down the benchmark for small SFs but for large SFs, the effect is more limited (maybe adding 9x0.5 min ~ 5 minutes in total).